I am joining Prescient to bring the wealth of insight offered by atomistic representations of protein structure to Prescient’s machine-learning toolkit. I served as a Research Assistant Professor at the University of North Carolina for the past 12 years having architected the Rosetta3 molecular modeling suite. While there, I developed new algorithms for protein specificity design that lead to the discovery of bispecific antibody mutations in both the Fab and Fc regions. I was a post-doc in the Baker lab at the University of Washington for two years and in the Kuhlman lab at the University of North Carolina for one year after earning my PhD in Computer Science at the University of North Carolina under Jack Snoeyink. I earned my BA from the University of Virginia.
I am bringing to the Prescient team deep experience in developing GPU-accelerated protein-structure-modeling software with the intention of incorporating the knowledge gained from decades of research in molecular-mechanics-force fields into the machine-learning models that are now driving the field of protein-structure prediction and protein design. Currently, the structures that are predicted by machine learning models such as AlphaFold frequently contain non-physical artifacts, such as chains passing through each other, with little insight offered to the user as to why the model would prefer such a structure. Ideally, the models should generate structures that not only capture the evolutionary history that MSAs reflect but that are at a low conformational energy, and I believe that incorporating atomistic energetics into the protein structure prediction pipeline will improve structure prediction and the interpretability of the results.
I have been closely involved with protein design projects for nearly twenty years and enjoy the thrill of designing new proteins with new functions. I am excited to work on more protein engineering projects in the near future.